An alternative way to classify missing data mechanism in clinical trials--a dialogue on missing data.

Area of Science:

• Pharmaceutical Statistics
• Clinical Trial Methodology
• Biostatistics

Background:

• Missing data due to patient dropout is a persistent challenge in pharmaceutical clinical trials.
• Defining the appropriate target population parameter for statistical inference with missing data remains a debated issue.
• Current methods often aim for hypothetical parameters or combine all dropout information, potentially misrepresenting drug effects.

Purpose of the Study:

• To propose a novel classification of patient dropouts in clinical trials.
• To differentiate between intrinsic (drug-related) and extrinsic (non-drug-related) dropouts.
• To establish a framework for defining a more accurate population parameter for treatment effect estimation.

Main Methods:

• Proposed classification of patient dropouts into intrinsic and extrinsic categories.
• Distinguishing dropouts based on their relationship to drug attributes versus non-drug-related factors (e.g., protocol deviation).
• Integrating this classification with existing missing data frameworks (MCAR, MAR, MNAR) for statistical analysis.

Main Results:

• Intrinsic dropouts (due to drug-related reasons) should inform the population parameter of treatment effect.
• Extrinsic dropouts (due to non-drug-related reasons) should not influence the target population parameter.
• This approach aims to provide a more accurate representation of a drug's true effect.

Conclusions:

• Classifying dropouts as intrinsic or extrinsic offers a more nuanced approach to handling missing data in clinical trials.
• This classification helps in determining a population parameter that fairly reflects the drug's efficacy and safety profile.
• Further considerations include statistical inference methods and practical implementation in real-world clinical trial settings.